Model-Based Drug Development - NEDMDG · PK/PD Modeling in Pediatric Drug Development Bernd...

New England Drug Metabolism Discussion New England Drug Metabolism Discussion Group, Group, Summer Symposium, June 14, 2011, Shrewsbury, MA

© Bernd Meibohm, PhD, FCP, University of Tennessee 1

New England Drug Metabolism Discussion GroupNew England Drug Metabolism Discussion GroupSummer Symposium, June 14, 2011, Shrewsbury, MA

The In Silico Child: PK/PD Modeling in Pediatric Drug Development

Bernd Meibohm, PhD, FCPBernd Meibohm, PhD, FCPProfessor & Associate Dean for Graduate Programs and ResearchProfessor & Associate Dean for Graduate Programs and Research

College of PharmacyCollege of PharmacyThe University of Tennessee Health Science CenterThe University of Tennessee Health Science Center

Memphis, TN, U.S.A. Memphis, TN, U.S.A.

A multi-disciplinary approach that integrates the relationships between diseases, drug characteristics, and individual variability

Model-Based Drug Development

y

A framework for synthesizing information and extrapolating beyond what is traditionally studied in RCTs A tool for rationale, critical decision makingFrom drug discovery to post-marketingA mathematical explanation of the relationships needed to explain clinical outcomes over a timeframe of interest at its

© Bernd Meibohm, PhD, FCP, University of Tennessee

explain clinical outcomes over a timeframe of interest at its coreAway from study centric approach: seamless data mining and knowledge management strategy that quantitatively integrates data across studies and development phases



DiseaseProgressionCompetitors

Patho-physiology Genetics

HealthcareDevelopment

Model-Based Drug Development

Model-basedInterdisciplinaryKnowledgebase

HealthcareOutcome

Development Tool Kit

Zhang, Pfister, MeibohmAAPS J 2008, 4:552-9

© Bernd Meibohm, PhD, FCP, University of TennesseeClinical DevelopmentPost-

marketing

Phase IPhase II

Phase III

App

rova

l

Preclinical Development

Development Continuum

Challenges in Pediatric Pharmacotherapy

Molecular & physiologic processes determining PK & PD undergo developmental changes based on a child’s maturational progress Different patients follow different developmental trajectories so that usual predictors such a PNA, PMA, or GA do not fully capture the variability in PK and PD related to developmental changesClinical studies in pediatric patients are challenging and limited due to ethical and logistic constraintsOff l b l th l th th th ti


Off-label use more the rule rather than the exceptionDevelopment of dosing recommendations

o Scaling from adult informationo Preclinical informationo Empirical: Anecdotal evidence/personal experience



Efficacy

Central Paradigm of Clinical Pharmacology

Dose Conc

Efficacy

T i it


Toxicity

Pharmacokinetics Pharmacodynamics

Conc

entr

ation

AUC 3AUC 2

Low exposure(~infant exposure)

AUC 1

Medium exposure(~adult exposure)

High exposure(~neonate exposure) Läer, Barrett, M

Determinants of Drug Response in Children

Time Time Time

fficac

y

Tox

icity

Meibohm

, J Clin Pharmacol 2009

Effe

ct (%)

1 2 3 1 2 31 2 3

Low sensitivity Medium sensitivity High sensivity

ExposureExposure Exposure


Ef Low sensitivity Medium sensitivity(~ adult sensitivity) High sensitivity

Low exposure(~ infant exposure)

No efficacyNo toxicity


Therapeutic efficacyNo toxicity

Medium exposure (~ adult exposure)




High exposure(~ neonate exposure)

Therapeutic efficacyToxicity



, 49, 889-904



Pediatric Study Decision Tree (FDA)


Meibohm et al., AAPS J 2005, 7, 475-87

Optimal drug response

The Three Pillars of Science

Hyp

othe

sis

Mod

ellin

g &

si

mul

atio

n

inic

al e

xper

imen

t


Therapeutic problem

Cl

Läer, Barrett, Meibohm, J Clin Pharmacol 2009, 49, 889-904



ModelingModels are simplified descriptions of certain aspects of reality by mathematical means, thereby allowing to concentrate on the factors believed to be important

Modeling & Simulation

pSummarizing measured data by integrating different measures and prior knowledge about biological processesIdentify the best model that sufficiently describes the data (Rule of Parsimony: simplest model)Purpose-driven: Level of model complexity defined by its intended use

Simulation


Modeling is a prerequisite for simulations: Application of the developed modelPredictions beyond the measured data: inter- or extrapolationsValidity of simulations depends on model (and the purpose is was developed for)Prediction error and uncertainty

Deterministic“Best guess” parameter point estimates used for simulationOne discrete outcome of simulation

Simulation ApproachesDeterministic vs. Stochastic

One discrete outcome of simulationo E.g. a discrete drug concentration vs. time profile

Parameters may be dependent on covariatesPro: Simplicity; ease of understandingCon: No uncertainty in parameter estimates considered

Stochastic (Monte-Carlo Simulations)Distributions for each specific parameter that capture the degree f t i t


p p pof uncertainty

o Repeated random sampling of parameters from these distributions to simulate the outcome based on the underlying structural model.

Distribution of outcomes with central tendency and spreadPro: provides inherently a measure of credibility and likelihood for simulation outcomesCon: increased complexity and thus difficult understanding and acceptance



Simulation ApproachesDeterministic vs. Stochastic

Discrete Concentration-Time Profile

Probability Distribution of Concentration-Time Profiles


Top Down ApproachesDeductive or Analytic Approaches

Based on generalized concepts in medical sciencesV f d d Verification-driven data mining process Allows expression of preconceived facts or theories in model termsTesting their validity within the context of the model system and the available dataIdentify reasons for the validation or invalidationConstant rebuilding and refinement of the model given the


g gresults of the continuous hypothesis testingOnly limited assumptions necessary

Data-driven process, in which the model is iteratively refined to optimally describe observed data



Top Down ApproachesDeductive Reasoning

DatasetDatasetTheory

Hypothesis testing


Model refinement

Confirmation

Current dosing guidelinesDifferent dosing regimens are proposed for vancomycin in pediatrics over last several years

o One standard dose for all neonates (DR1)

VancomycinClinical Issues in Pediatrics

o One standard dose for all neonates (DR1)o PMA based dosing (DR2)o PMA and PNA based dosing (DR3)o Serum creatinine based dosing (DR4)

No consensus among clinicians on the use of any standard vancomycin dosing regimen in term and preterm neonates

o Comparative evaluation in clinical trial hampered by logistic and ethical constraints

Little information on dosing in premature vs. term neonateso Growing demand based on more sophisticated NICUs


Gr w ng man a n m r ph t cat NInitial dosing frequently followed by a posteriori individualization based on Sawchuk-Zaske method or Bayesian forecastingComparative clinical study nearly impossible due to logistic and ethical constraints of studies in neonates

ObjectivesTo evaluate the four standard dosing regimens in silico for their ability to achieve target vancomycin concentrations



Retrospective non interventional clinical study

Step 1: PopPK AnalysisEstablish pharmacostatistic vancomycin disposition model based on local population

Retrospective, non-interventional clinical study LeBonheur Children’s Medical Center, Memphis, TN

Inclusion criteria: Full term and premature neonates with PMA ≤44 wks At least one vanc serum concentration levelDocumented vanc dose, dosing schedule, and time of blood draw

Exclusion criteria: Concurrent nephrotoxic drugs (i.e., amphotericin B, cisplatin)Extracorporeal membrane oxygenation or hemodialysis


Extracorporeal membrane oxygenation or hemodialysisCovariates tested:

Gestational age (GA), postnatal age (PNA), postmentrual age (PMA), weight, serum creatinine, blood urea nitrogen, and urine output.

N Male/ Female Preterm/Term Weight GA PNA SCR134 54%/ 46% 66%/ 34% 0.64-5.3 kg 23-41 wks 1-121 days 0.2-2.5 mg/dL

Parameter ModelEstimate (%RSE)

Step 1: PopPK AnalysisNonlinear Mixed Effects Modeling

Parameter Model (%RSE)

CL (L/hr) θ1*(WT/2.5)0.75*(SCR/0.42) θ3*(PMA/37) θ4 θ1 0.18 (3.5)θ3 0.7 (9.0)θ4 1.41 (15.4)

V (L) θ2*(WT/2.5) θ2 1.7 (7.3)

BSV CL Between subject variability in Clearance 25% (18.7)

BSV VBetween subject variability in Volume of Distribution 22% (51 7)


Parameters allometrically scaled for weight using exponent of ¾ for CL and 1 for VdSCR and PMA explained 53% of the BSV in CL

BSV V of Distribution 22% (51.7)

Residual errorProportional:16.4% (36.7%)

Additive:1.5 µg/ml (37.5%)



Monte-Carlo simulations based on PopPK modelEach dosing regimen was treated as a different scenario having

Step 2: Simulations of DRs

scenario havingo 200 replicates per scenarios o 100 subjects per replicateMultivariate age-weight distributions

o Term Neonates: CDC growth chartso Preterm neonates: intra-uterine and postnatal growth charts Covariance between PNA and SCR based on original


Covariance between PNA and SCR based on original dataset

Outcome metricso Trough concentrations 0.5 hr prior to next doseo Percentage of subjects having trough concentration between

5-15 µg/ml

Multivariate Covariate Distribution

Premature TermTransitional

Physiologically reasonable covariate vectors

PrematureGA 20-34 wks

TermGA 39-42 wks

TransitionalGA 35-38 wks

PNA>7d PNA7d PNA



Virtual Patient Population

5

6

PREM 5, PNA>7PREM 50, PNA>7PREM 95, PNA>7

Wei

ght (

Kg)

2

3

4

5PREM 5, PNA7TRANS 5, PNA



Bottom Up ApproachesInductive or Synthetic Approaches

Synthesize individual pieces of information to form broader generalizations and theoriesg

o Observationso Datao Patterns

Individual elements are linked together to form larger subsystems based on Systems Biology

Subsystems are linked via many levels to from a complete top-


y y p plevel modelAssumption-rich approachHypotheses are driven by observed patterns and interdependenciesPredictions possible without experimental (clinical) data

Confirmation

Inductive Reasoning

Bottom Up Approaches

Theory

Hypothesis testing


Subsystems

Data – Observations - Patterns



Example: Physiologic PK Modeling


Integration of information from multiple sourcessources

Types of data:o Anatomical&Physiologicalo Pathophysiologicalo Drug-specifico Molecular

Relevant organ & tissue function individually

d d


consideredBased on ‘typical’ parameters for specific subpopulationUncertainty may be implementedKuepfer, Mol Syst Biol 2010, 6, 409Rowland et al., Annu Rev PharmacolToxicol 2011, 51, 45-73

Developmental Changes in

Kearns et al. N Engl J Med 2003, 349:1157-67

gPhysiologic Factors

Affecting Drug Disposition

© Bernd Meibohm, PhD, FCP, University of Tennessee24



Koukouritaki et al., J Pharmacol Exp Ther 2004, 308:965-74/m

g) CYP2C9

Ontogeny of Hepatic DME (I)CYP2C

CYP2

C9 (pm

ol/

mg)

CYP2C19


CYP2

C19

(pmol/m

Ontogeny of Hepatic DME (II)Hepatic Phase I & Phase II Enzymes


Alcorn & McNamara, Clin Pharmacokinet 2002, 41, 959-98

Strassburg et al., Gut 2002, 50, 259-62



Ontogeny of Hepatic DME (III)Modeling of Developmental Trajectories


Alcorn & McNamara, Clin Pharmacokinet 2002, 41, 1077-94

Example: CYP2E1 & Toluene ClearanceCYP2E1 content vs. age

PBPK-model

Toluene intrinsic clearance vs. ageToluene hepatic

clearance vs. CYP2E1

© Bernd Meibohm, PhD, FCP, University of Tennessee28Nong et al., Toxicol Appl Pharmacol 2006, 214, 78-87



The In Silico ChildDual Direction Up & Down Approach

Top Down Approaches



Dosing Recommendations & Dosage Individualization

The In Silico ChildDual Direction Up & Down Approach

Usually not sufficient data available to describe full b tt d l tbottom up model system

Dependence on assumptions with associated uncertaintyIncorporation of bottom up subsystem components in top down methodology to explain residual variability

Multi-scale models based on systems biology


Combination of both methodologies is the most promising approach to give birth to the In SilicoChild



Example #1Example #1

Inhalational Anthrax in ChildrenLevofloxacin therapy

Rare bacterial infection with B. anthracisInhaled spores germinate in lower respiratory tract, leading to toxin-p g m p y , gmediated necrotizing pneumonia with respiratory failure and deathBioterrorism threatEfficacy studies not feasible

Fluoroquinolones as empirical therapy Despite usually limited use in children due to lesions in the cartilage of juvenile animalsPotentially life-saving treatment outweighs risks

Development of an optimized pediatric dosing regimen for


Development of an optimized pediatric dosing regimen for children from 6 months to 17 years

Based on exposure in animal experiments that prevents progression of pulmonary anthrax after inhalation exposureExtrapolating adult data to children

o No difference in C-R relationshipo Beneficial and adverse drug effects similar between children and adultso Only adjustment for PK differences



PK characteristicso Linear PK

99% l b l b l

Inhalational Anthrax in ChildrenLevofloxacin Population Pharmacokinetics

o 99% oral bioavailabilityo Renal excretion as major elimination pathway (fe = .87)

Development pharmacologyo Disposition as function of body sizeo Maturation of renal function within first 2-3 yrs

Available PK datao 90 pediatric patients (7 mg/kg) and 47 healthy adults (500 or 750 mg)


CL ~ BW0.43 x (Age/[Age+A50]); A50=0.34 yrAllometrically weight-adjusted clearance with maturation function for renal eliminationRenal function maturation: 60% for 6-month old; 75% for 1-yr oldStrictly limited to lower age range of 6 monthsV ~ BW

Li et al., Antimicrob Agents Chemother 2010, 54, 375-9

Inhalational Anthrax in ChildrenMaturation of Renal Excretion Function

Estimated based on Levofloxacin PopPK Analysis

Creatinine Clearance vs. Postnatal Age p y

25

75

125

eaCL

[mL/

min/1

.73

m2 ]

+2 SD

-2 SD

g



Age [years]1 2 3 4 Normal

adult00

25

Cre

Holliday et al., Pediatric Nephrology, Williams & Wilkins 1994



Simulation of exposuresTo match adult exposures for 500 mg QD with AUC24, Css,max, Css,min15 mg/kg QD in children matches adult 500 mg QD

Inhalational Anthrax in ChildrenPediatric Dosing Recommendations

15 mg/kg QD in children matches adult 500 mg QDCmax,ss exceeds adult values: 7.5 mg/kg BID dosingFor children >50 kg: 500 mg QD (adult dose)


Final recommendation8 mg/kg BID (500 mg QD for children >50 kg)


Example #2Example #2



Sotalol in Pediatric Supraventricular Tachycardia

Sotalol Plasma Concentrations in a 5 month-old (5.5 mg/kg; open circles; 7.8 mg/kg; closed circles) and a 26 yr-old Patient (2 mg/kg)


Läer et al., Eur J Clin Pharmacol 2001, 57, 181-2

Sotalol in Pediatric PatientsPharmacokinetic Properties

BCS class 1 (hydrophilic; secondary amine)

N (%) Mean (SD) Range

A [ ] 82 3 0 (4 2) 0 03 13 7

Study Population

amine)Oral bioavailability 90-100 %Linear PKNo plasma protein bindingElimination

Terminal half-life: ~12 hr (adults)80-90% renal excretion in

h d f

Age [yrs] 82 3.0 (4.2) 0.03 - 13.7

Newborns 14 (17) 0.050 (0.015) 0.03 - 0.07

Infants 38 (46) 0.53 (0.43) 0.09 - 1.7

Children 30 (37) 7.6 (3.7) 2.1 - 13.7

Sex (male) 51 (62)

Weight [kg] 14.3 (15.8) 2.2 - 84.3

Height [cm] 84 (36) 35 - 178


unchanged formEssentially no metabolismRenal CL > GFR (~1.5 x) and reduced by cimetidine coadministration

filtration and secretion

BSA [m2] 0.55 (0.41) 0.17 - 2.04

SCr [mg/dL] 0.44 (0.15) 0.2 - 1.0

GFR [mL/min] 42.8 (45.6) 3.1 - 232

NGFR [mL/min/1.73m2] 103 (45.5) 31.5 - 217



Age-Specific PharmacokineticsCL ~ BW0.75 x (1+S x Age0.142); S=1 for Age



Sotalol – Pediatric C/R

EC50: ~0.45 µg/mLEC95: ~1.00 µg/mL

Filled circles = neonates

© Bernd Meibohm, PhD, FCP, University of Tennessee41Läer et al., J Am Coll Cardiol 2005, 46, 1322-30

Age-Specific Pediatric Dosage RecommendationsBased on Exposure-Response Relationship

© Bernd Meibohm, PhD, FCP, University of Tennessee42Läer et al., J Am Coll Cardiol 2005, 46, 1322-30



The FutureThe In Silico Child in Drug Development & Clinical Practice

Integrated M&S using all available sources of information

Methodologiesavailable sources of information

Molecular & Mechanistic InformationAnatomy, Physiology and PathophysiologyOntogenyClinical observations and h b b

Bottom up & Top DownBayesian priorsIndividual vs. populationMonte Carlo simulations as coreEasy user interface for widespread acceptance


their between-subject variabilityHistoric dataComparator compoundsClass effectsSpecies & subpopulations

widespread acceptanceEducation, education, education…


Model-Based Drug Development - NEDMDG · PK/PD Modeling in Pediatric Drug Development Bernd...

Documents

Transcript of Model-Based Drug Development - NEDMDG · PK/PD Modeling in Pediatric Drug Development Bernd...